Professional Documents
Culture Documents
Article
Development of a Scale-up Tool for
Pervaporation Processes
Holger Thiess, Axel Schmidt ID
and Jochen Strube *
Institute for Separation and Process Technology, Clausthal University of Technology, Leibnizstraße 15,
38678 Clausthal-Zellerfeld, Germany; thiess@itv.tu-clausthal.de (H.T.); schmidt@itv.tu-clausthal.de (A.S.)
* Correspondence: strube@itv.tu-clausthal.de; Tel.: +49-532-372-2200
Abstract: In this study, an engineering tool for the design and optimization of pervaporation processes
is developed based on physico-chemical modelling coupled with laboratory/mini-plant experiments.
The model incorporates the solution-diffusion-mechanism, polarization effects (concentration and
temperature), axial dispersion, pressure drop and the temperature drop in the feed channel due to
vaporization of the permeating components. The permeance, being the key model parameter, was
determined via dehydration experiments on a mini-plant scale for the binary mixtures ethanol/water
and ethyl acetate/water. A second set of experimental data was utilized for the validation of the model
for two chemical systems. The industrially relevant ternary mixture, ethanol/ethyl acetate/water,
was investigated close to its azeotropic point and compared to a simulation conducted with the
determined binary permeance data. Experimental and simulation data proved to agree very well
for the investigated process conditions. In order to test the scalability of the developed engineering
tool, large-scale data from an industrial pervaporation plant used for the dehydration of ethanol was
compared to a process simulation conducted with the validated physico-chemical model. Since the
membranes employed in both mini-plant and industrial scale were of the same type, the permeance
data could be transferred. The comparison of the measured and simulated data proved the scalability
of the derived model.
1. Introduction
Extractive, pressure-swing and azeotropic distillation are typically applied as state-of-the-art
unit operations for the separation of azeotropic mixtures, despite their technical complexity and
large energy consumption [1,2]. Optimizing the energy efficiency of new and established processes
becomes more and more important in a globalized market [3]. Hence, pervaporation and vapour
permeation processes were developed as stand-alone or hybrid processes as alternatives to distillation
unit operations. Pervaporation is a unit operation through which a liquid mixture is selectively
separated by a membrane. The driving force of the process is the partial pressure difference between
the feed and permeate side of the membrane, instead of the difference in volatility on which distillation
is based. Thereby, a pervaporation process is able to separate azeotropic mixtures without the use of
entrainers or a variation in pressure [4].
The design and optimization of pervaporation processes as well as the possible interconnection
with other unit operations necessitates a high theoretical understanding of the process, membrane
and feed mixture, thus leading to a demand for proper engineering tools for designing pervaporation
unit operations.
The scope of this work is the development of a physico-chemical model which, if combined
with experiments for the model parameter determination, can be used as an engineering tool for the
scale-up and optimization of pervaporation processes. The physico-chemical model includes balances
for mass, enthalpy and impulse. It incorporates the transport phenomena known in literature such as
the solution-diffusion mechanism, concentration and temperature polarization, axial dispersion and
pressure drop in open channels as well. In order to determine the permeance and to validate the model,
dehydration experiments were conducted with the two binary chemical systems ethanol/water and
ethyl acetate/water. The experimental design incorporates variations in mixture composition, feed
temperature, permeate pressure and feed volume flow. The ternary system ethanol/ethyl acetate/water
was investigated at two temperatures for a feed composition close to the azeotropic point as well.
With the derived binary permeance data, simulations are compared to experimental data
of the industrially relevant ternary mixture ethanol/ethyl acetate/water. The scalability of the
physico-chemical model was investigated by comparing data from an industrial pervaporation plant
(200 m2 , ethanol/water) with the simulation of the derived model and permeance data.
2. Theory
In order to derive a consistent physico-chemical model, a defined control volume combined
with balancing equations describing mass, heat and impulse transfer is essential. In the following
the fundamental equations used for mass, enthalpy and momentum balancing as well as the model
parameter determination methods are introduced.
dµi
Ji = − Li , (1)
dz
with Li being a proportionality coefficient and µi the chemical potential. Restricting to molar
concentration and pressure gradients, the chemical potential is written as [5]:
with the gas constant R, the temperature T, the component’s activity coefficient γi , the molar fraction
xi , the molar volume vi and the pressure p. For a dense membrane, the pressure inside the membrane
is constant and equal to the pressure in the feed channel. Under this premise combining Equations (1)
and (2) results in Equation (3)
dln(γi xi ) RT dxi
Ji = − Li RT = − Li . (3)
dz xi dz
Membranes 2018, 8, 4 3 of 27
Similar to Fick’s law the proportionality term is replaced by the diffusion coefficient Di
(Equation (4)), resulting in Equation (5).
RT
Di = L i (4)
xi
x i0( m ) − x i l ( m )
Ji = Di (5)
l
Therefore, the difference in molar concentration over the membrane thickness has to be
investigated in further detail. The integration of Equation (2), while using the vapour pressure
pisat as pressure reference, results in Equation (6) for incompressible and Equation (7) for compressible
fluids, respectively.
µi = µ0i + RT ln(γi xi ) + vi ( p − pisat ) (6)
p
µi = µ0i + RT ln(γi xi ) + RT ln (7)
pisat
The following two assumptions are made [5]:
The feed side of the membrane can thus be expressed by Equation (10)
µ0i + RT ln γiL0 xiL0 + vi ( p0 − pisat ) = µ0i + RT ln γi0(m) xi0(m) + vi ( p0 − pisat ), (10)
γiL0 xiL0
x i0( m ) = = Ki · xiL0 , (11)
γi 0 ( m )
where Ki is the liquid phase sorption coefficient. In analogy to the retentate, the permeate side is
expressed with Equation (12):
pl
µ0i + RT ln γiGl xiGl + RT ln = µ0i + RT ln γil (m) xil (m) + vi ( p0 − pisat ). (12)
pisat
γiG pl
−vi ( p0 − pisat )
l G
xil (m) = · · x · exp , (13)
γi l ( m ) pisat il RT
where the exponential term is near one, thus resulting in Equation (14)
γiG pil
xil (m) = l
· = KiG · pil , (14)
γi l ( m ) pisat
with the gas phase sorption coefficient KiG . In order to work with just one type of sorption coefficient,
a hypothetical gas phase between the liquid phase and the membrane is introduced (Figure 1).
Membranes 2018, 8, 4 4 of 27
Membranes 2018, 8, 4 4 of 27
Figure1.1. Hypothetical
Figure Hypothetical gas
gas phase
phase between
between feed
feed solution
solution and
and membrane.
membrane. Illustration
Illustrationadapted
adaptedfrom
from[6].
[6].
Thereby, the equilibrium between the liquid and the hypothetical gas phase can be expressed by
Thereby, the equilibrium between the liquid and the hypothetical gas phase can be expressed by
Equation (15)
Equation (15)
+ + − = + +
p0 ,
(15)
L L G G
µ0i 0
+ RT ln γi0 xi0 + vi ( p0 − pisat ) = µi + RT ln γi0 xi0 + RTln , (15)
Pisat
where rearranging leads to Equation (16)
where rearranging leads to Equation (16)
= (16)
γ⋅ iG
x i0 = L 0 p i0 (16)
Implementing Equation (16) in Equation (11) · pisat in
γi results
0
µil (m=
)
= µ( il ) (21)
(20)
= ⋅ µ i0 =
⋅ µi0⋅(m) − ⋅ (21)
(22)
Ji = Qi · pi sat · xiL0 · γiL0 − p p · yil (22)
2.1.2. Determination of the Permeance
2.1.2.In
Determination of the approaches
literature, several Permeance for the determination of the permeance exist. Since the
effects sorption and diffusion are dependent on concentration and temperature, it becomes clear,
In literature, several approaches for the determination of the permeance Qi exist. Since the that
effects
the permeance
sorption must show
and diffusion this dependency
are dependent as well. For
on concentration and short-cut models,
temperature, describing
it becomes isothermal
clear, that the
processes without a larger change in concentration, the permeance can be assumed constant
(Equation (23)), to obtain a quick prediction of the flux with just few experiments [7].
Membranes 2018, 8, 4 5 of 27
permeance must show this dependency as well. For short-cut models, describing isothermal processes
without a larger change in concentration, the permeance can be assumed constant (Equation (23)),
to obtain a quick prediction of the flux with just few experiments [7].
Qi = constant (23)
The reference permeance Q0i and the permeance parameter Bi , which is in literature often
referred to as the activation energy in literature [2], based on the reference temperature T0 are
pseudo-physical parameters, which can be determined using experimental data. In order to implement
a concentration dependency, different approaches are known, based on the applied membranes
and solvents. Equations (25) and (26) show different modifications of the Arrhenius approach,
to incorporate the water concentration’s influence [2,8] on the permeance for hydrophilic membranes.
B 1 1
Qi = Q0i · exp Ai w H2 O − i − (25)
R T0 T
A B 1 1
Qi = Q0i · w Hi2 O · exp − i − (26)
R T0 T
In some cases, no temperature dependency for the permeance is observed. This indicates a
compensation between the exothermal sorption and the acceleration of the diffusion process, resulting
in a temperature-independent permeance [9].
By integrating the diffusion and solubility coefficient into the permeance, all coupling phenomena
based either on solubility or diffusivity of the components are merged. The utilization of the Arrhenius
approach and the structure of Equations (24)–(26) identify the determination of the permeance as
clearly semi-empirical.
The ability to separate a fluid mixture is commonly described using separation factors or the
selectivity. The binary selectivity of a pervaporation process is defined in Equation (27) by the
permeance ratio of the more permeable component over the less permeable [10]:
Qi
αij = . (27)
Qj
dwi
Ji, Di f f = −ρtotal,F · Di , (30)
dz
Ji, TM = Jtotal · wi,P , (31)
with the diffusion coefficient Di , the the feed mixture density ρtot, f and the total permeate flux Jtot
dwi
Jtot · wi + ρtot, f · Di = Jtot · wi,P (32)
dz
Integrating Equation (32) with the given boundaries shown in Equation (33) and rearranging
leads to the concentration polarization Equation (34).
wi =wi, f zZ=δ
dwi Jtot
Z
=− dz (33)
wi − wi,p ρtot, f · Di
wi =wi, f m z =0
!
Di wi, f m − wi,p
Jtot = ρtot, f ln (34)
δ wi, f − wi,p
The indices FM, F and P stand for the membrane surface on the feed-side, the bulk and the
permeate respectively. The ratio of the diffusion coefficient to the boundary layer thickness δ can be
superseded by the mass transfer coefficient k i (Equation (35)).
Di
ki = (35)
δ
The mass transfer coefficient may be calculated by the Sh-correlation shown in Equation (36),
based on the dimensionless quantities Re (Reynolds number), Sc (Schmidt number) and Sh (Sherwood
number). The definitions of the Re and Sc are given in Equations (37) and (38). The parameters a, b, c
and d are listed on literature based on different membrane modules and flow regimes [12–16].
a4
k dh dh
Sh = = a1 · Re a2 · Sc a3 · (36)
Di l
ρ u dh
Re = (37)
η
η
Sc = (38)
ρ·D
The concentration polarization influences the flux for pervaporation processes in two ways.
The more retained component accumulates at the membrane surface, thereby increasing its partial
pressure difference (see solution-diffusion model), hence its permeate flux. The less retained component
depletes at the membrane surface, resulting in a smaller permeate flux. This phenomenon is shown in
Figure 2. Both effects lead to a worse selectivity of the pervaporation process. An increase of the mass
transfer coefficients hones the diffusional transport and hence the selectivity of the process.
Membranes 2018, 8, 4 7 of 27
Membranes 2018, 8, 4 7 of 27
Figure 2. Illustration of the concentration polarization effect for an accumulating substance (above)
Figure 2. Illustration of the concentration polarization effect for an accumulating substance (above)
and aa depleting
and depleting substance
substance (below).
(below). Adapted
Adapted from
from [17].
[17].
∂mi = . , − . , + . , − . , −. , (39)
= mi,conv z − mi,conv z+dz + mi,disp − mi,disp − mi,PV (39)
∂t z z+dz
Each term can be rewritten, as shown in Equations (40)–(43):
Each term can be rewritten, as shown in Equations (40)–(43):
= ( ∙ )= ∙ ∙ = ∙ ∙ . (40)
∂mi ∂ ∂ ∂ci
= (ci ·dV ) = c · A ·dz = · A ·dz. (40)
∂t ∂t ∂t i , Q ∂t Q
, − , ≈ − . ⋅ = − ⋅ ⋅ ⋅ (41)
. . ∂ m i,conv ∂c i
mi,conv z − mi,conv z+dz ≈ − · dz = −u · AQ · · dz (41)
,
∂z ∂z
, − ,. ≈− ⋅ = ⋅ , ⋅ ⋅
.
. ∂mi,disp ∂ ∂ci ∂2 c (42)
mi,disp − mi,disp ≈− · dz = AQ · Dax,i · · dz = AQ · Dax,i · 2i · dz (42)
z z+dz =
∂z ⋅ , ⋅
∂z ⋅ ∂z ∂z
.
. ∂mi,PV
mi,PV = · dz = Ji · dA = Ji · wm · dz. (43)
= ∂z , ⋅ = ⋅ = ⋅ ⋅ . (43)
,
In order to work with mass fractions, Equations (40)–(43) are set against Equation (44) resulting in
In order
Equation (45).to work with mass fractions, Equations (40)–(43) are set against Equation (44) resulting
in Equation (45). mtot |z+dz = ρtot · dV = ρtot · AQ · dz = ρtot · (wm hch ) · dz (44)
| ∂w= i
⋅ ∂w =i ⋅ ⋅∂2 w= i
⋅ J(i ℎ )⋅ (44)
= −u · + Dax,i · 2
− (45)
∂t ∂z ∂z ρtot · hch
=− ⋅ + , ⋅ − (45)
⋅ℎ
Membranes 2018, 8, 4 8 of 27
The enthalpy flow of the feed solution can be derived by the mass flow, the specific heat capacity
of the feed as well as its temperature (Equation (47)):
. .
H f = m f · cep, f ( TF − T0 ). (47)
With the simplification that the temperature in feed and permeate is the same, Equation (48)
shows that the enthalpy flow of the permeate is based on the permeate flux, the membrane area and
the average specific enthalpy of the permeate.
.
hG
Hp = e eG
p · Jtot · Am = h p · Jtot · wm · lm (48)
To describe the change in enthalpy over the length ∆z, Equation (49) is derived:
. .
m f · cep, f T f − T0 | = m f · cep, f T f − T0 | hG
+e p · Jtot · wm · ∆z |z . (49)
z z+∆z
The differential change in the enthalpy flow for feed- and permeate-side of the membrane can
thereby be determined by:
.
d H r (z)
hG
= − Jtot (z) · wm · e p, (50)
dz
.
d H p (z)
hG
= Jtot (z) · wm · e p. (51)
dz
hV
∆e eG e L
tot = h p − h p (52)
2.2.2. Temperature Polarization
During the pervaporation process, the permeating components desorb and vaporize on the
permeate side of the membrane. Thus, vaporization enthalpy ∆ℎ is needed and withdrawn from
the feed solution
Membranes 2018, 8, 4 (see Figure 3). The overall vaporization enthalpy required is the enthalpy difference
9 of 27
between the gaseous and liquid phase, shown in Equation (52).
Figure 3. Illustration of the heat transfer via vaporization. Adapted from [21].
Figure 3. Illustration of the heat transfer via vaporization. Adapted from [21].
Similar to the effect of concentration polarization the developed boundary layer on the membrane
surface causes a temperature gradient between the bulk and the membrane surface on the feed side.
Hence, similar to concentration polarization the effect of temperature polarization decreases the
driving force and thereby the permeate flux. The enthalpy flow needed to vaporize the permeate mass
.
flow H vap is connected to the induced temperature gradient by the heat transfer approach shown in
Equations (53) and (54) [21]:
. vap
H vap = Jtot · A · ∆e
htot = Jtot · A · ehG
p − h
e L
p , (53)
.
H vap λ
= Tf − Tf m . (54)
Am δ
Analogous to Section 2.1.3 the thermal conductivity coefficient and boundary layer thickness are
combined into the heat transfer coefficient, given by Equation (55)
λ
α= (55)
δ
The Sherwood correlation introduced in Section 2.1.3 is based on the Nusselt correlation for heat
transfer. Sherwood and Schmidt number represent the mass transfer analogies to Nusselt and Prandtl
number in heat transfer. The definitions of the Nu and Pr are given by Equations (56) and (57).
a4
α · dh
dh
Nu = = a1 · Re a2 · Pr a3 · (56)
λ l
η · cep
Pr = (57)
λ·M
Rearranging Equation (58) to describe the differential change in pressure drop results in:
dp ρ u2
= −ζ . (59)
dz 2 dh
Depending on the flow regime, the drag coefficient ζ for open channels in plate and frame modules
can be expressed by Equation (60) for laminar flow (Re < 2320) and Equation (61) for turbulent flow
(Re > 2320) [21].
38
ζ= (60)
Re
1.22
ζ= (61)
Re0.252
Implementing Equation (60) for laminar flow and Equation (61) for turbulent flow into
Equation (58) leads to:
dp u
= −19 η 2 (62)
dz dh
dp ρ0.748 u1.748
= −0.61 η 0.252 . (63)
dz d1.252
h
Since the feed solution can be considered as an incompressible fluid, the velocity can be expressed
with the continuity equation for both laminar and turbulent flow regime:
.
dp f Vf
= −19 ηF 2 (64)
dz dh, f wm hch, f
. 1.748
dp f ρ0.748
f Vf
= −0.61 η 0.252
f . (65)
dz dh, f wm h1.748
1.252 1.748
ch, f
In case of the vapour on the permeate side, the ideal gas law is utilized since the permeate pressure
is adequately low, shown in Equation (66).
. .
R TP n p R Tp n p
uP = = (66)
p p AQ p p wm hch,p
Implementation of Equation (66) in Equations (62) and (63) leads to Equations (67) and (68)
which are used to describe the discretized pressure drop of the permeate for laminar and turbulent
flow regimes.
.
dp p R Tp n p
= −19 η p (67)
dz p p wm hch,p d2h,p
.
1.748
R Tp n P
dp p p p wm hch,p
= −0.61 η 0.252
p ρ0.748
p (68)
dz d1.252
h,p
For the sake of completeness, it has to be stated that between Re 2000 and Re 4000 a transition
state exists and even within the turbulent flow regime the nature of the flow can be further divided.
algorithm. The investigated pervaporation processes were batch processes. Therefore, the model
consists of three submodels: the feed tank, the membrane module and a permeate tank representing
the condensed permeate. In the pervaporation plants as well as the model, the feed leaves the feed
tank and enters the module. The permeate stream leaves the module and accumulates in the permeate
tank. The retentate stream is redirected into the feed tank. In both feed and permeate tank a simple
mass balance was implemented. The initial mass of each component, the temperature and the overall
feed mass flow and the pressure are needed for the feed tank submodel. The permeate tank submodel
requires the permeate pressure. Both, feed and permeate tank are assumed to be ideally mixed.
The initial masses of the components in the permeate tank are zero.
Figure 4 depicts the development of the rigorous pervaporation model used for the membrane
module in this work. Based on the shown control volume the model includes balances for mass,
enthalpy and impulse. The overall control volume is divided into five sub-control volumes. CV1 is the
feed channel of the membrane module without the boundary layer on the membrane surface. 11 of 27
Membranes 2018, 8, 4
Figure 4. Left: Course of action used for the pervaporation model development. Right: Overview of
Figure 4. Left: Course of action used for the pervaporation model development. Right: Overview of
the five control volumes used in modelling and simulation. Adapted from [23].
the five control volumes used in modelling and simulation. Adapted from [23].
The mass balance to describe the bulk flow is based on the DPF model (Equation (45)). As
The
boundarymass balance tofor
conditions describe the bulk
entrance and flow
exit isofbased
the on thechannel
feed DPF model (Equation(69)
Equations (45)).
andAs boundary
(70) were
conditions for
implemented: entrance and exit of the feed channel Equations (69) and (70) were implemented:
.
m f ,i ,
wr,i |, Z=0 == . ,, (69)
(69)
,
m f ,tot
dwr,i,
== 0.0. (70)
(70)
dz z= L
Equation (50)
Equation (50) was
was utilized
utilized to
to describe
describe the
the enthalpy
enthalpy balance
balance in
in the
the feed
feed channel
channel and
and its
its exit
exitin
in
CV1. The entrance boundary condition of the feed channel is based on the feed temperature,
CV1. The entrance boundary condition of the feed channel is based on the feed temperature, the feed the feed
compositionand
composition andthe
thefeed
feedmass
massflow:
flow:
. =. = . , ⋅ ̃ , − . (71)
Hr = H f = m f ,tot · cep, f T f − T0 . (71)
As impulse balance in CV1Z=Equation
0
(64) combined with Equations (72) and (73) were
incorporated into
As impulse the model.
balance in CV1 Equation (64) combined with Equations (72) and (73) were incorporated
into the model. | = (72)
p r | z =0 = p f (72)
= 0 (73)
CV2 represents the boundary layer on the feed side of the membrane. Concentration and
temperature polarization effects occurring in CV2 were implemented with Equations (34) and (54)
respectively. The mass transfer through the membrane in CV3 was described by the LDM (Equation
(22)). CV4 was neglected in this model. Equation (67) coupled with the boundary conditions shown
Membranes 2018, 8, 4 12 of 27
dpr
=0 (73)
dz z= L
CV2 represents the boundary layer on the feed side of the membrane. Concentration and
temperature polarization effects occurring in CV2 were implemented with Equations (34) and
(54) respectively. The mass transfer through the membrane in CV3 was described by the LDM
(Equation (22)). CV4 was neglected in this model. Equation (67) coupled with the boundary conditions
shown in Equations (74) and (75) were used to describe the pressure drop in CV5:
dp p
= 0, (74)
dz z=0
p p z= L = p p_exp (75)
For the description of the mass transfer in CV5, a simple balancing of the streams was (Equation (76))
used together with the boundary condition shown in Equation (77):
.
dm p,i
= Ji · wm , (76)
dz
.
m p,i z=0 = 0 (77)
AQ wm · hch
dh = 4 =2 . (78)
Uwetted wm + hch
With the known material data and the hydraulic diameter, the Sherwood (Equation (36)) and
Nusselt correlation (Equation (56)) were utilized to obtain the mass transfer and heat transfer coefficient.
The required parameters a1 to a4 were obtained from literature data for the respective flow regime and
listed in Table 1.
Flow Regime a1 a2 a3 a4
Laminar (Re < 2300) 1.615 0.33 0.33 0.33
Turbulent (Re > 2300) 0.026 0.80 0.30 0
The axial dispersion coefficient Dax for Reynolds numbers smaller than 10,000 in open channels
was correlated by Equation (79) [19]:
Dax 1 Re·Sc
= + . (79)
uL Re·Sc 192
Figure 5. Flowsheet of the pervaporation unit utilized for the binary and ternary experiments.
Figure 5. Flowsheet of the pervaporation unit utilized for the binary and ternary experiments.
Table 2. Experimental design for the binary system ethanol/water. Indicated (*) experiments represent
Table 2. Experimental design for the binary system ethanol/water. Indicated (*) experiments represent
the centre point experiments. Bold highlighted experiments were used for the determination of the
the centre point experiments. Bold highlighted experiments were used for the determination of the
permeance, the rest for the validation of the model. All experiments were conducted with feed
permeance, the rest for the validation of the model. All experiments were conducted with feed pressure
pressure of 3.5 bar.
of 3.5 bar.
Exp. , (%) (°C) (mbar) (L/h)
. , (%)
Exp. wstart
f,ethanol (%) Tf (◦ C) pp (mbar) Vf (L/h) wend
f,ethanol (%)
1 84.6 55 10 40 84.8
1 2 84.6
85.3 55
75 10
100 70 40 87.4 84.8
2 85.3 75 100 70 87.4
3 3 85.2
85.2 95
95 1010 70 70 96.4 96.4
4 4 86.2
86.2 95
95 100
100 40 40 94.6 94.6
5*5* 90.1
90.1 85
85 5555 55 55 93.8 93.8
6* 90.0 85 55 55 94.5
7* 90.3 85 55 55 94.7
8 94.9 75 10 70 96.6
9 95.0 75 100 40 95.3
10 95.4 95 10 40 98.6
11 95.2 95 100 70 97.1
Table 3. Experimental design for the binary system ethyl acetate/water. Indicated (*) experiments
represent the centre point experiments. Bold highlighted experiments were used for the determination
of the permeance, the rest for the validation of the model. All experiments were conducted with feed
pressure of 3.5 bar.
.
Exp. wstart
f,ethyl acetate (%) Tf (◦ C) pp (mbar) Vf (L/h) wend
f,ethyl acetate (%)
1 97.1 50 10 40 98.2
2 97.2 50 100 70 97.3
3 96.6 70 10 70 97.8
4 97.0 70 100 40 98.4
5* 97.2 60 55 55 97.6
6* 98.1 60 55 55 98.7
7* 97.9 60 55 55 98.4
8 99.1 70 10 40 99.1
9 92.3 95 10 40 99.1
10 93.9 95 100 70 99.0
11 92.5 70 100 40 98.2
12 91.7 70 10 70 98.0
Membranes 2018, 8, 4 15 of 27
Table 4. Overview of process parameters used in the ternary experiments. The mass fractions are given
for the start of the experiments and in brackets after 10 h. The mass fractions listed are the mean value
based on the thrice-repeated experiments. All experiments were conducted with a feed pressure of
5 bar.
.
Exp. wf,ethanol (%) wf,ethyl acetate (%) wf,H2 O (%) Tf (◦ C) pp (mbar) Vf (L h−1 )
1 16.0 (16.4) 76.1 (81.3) 7.9 (1.4) 95 54 70
2 16.0 (17.1) 75.7 (78.1) 8.3 (4.1) 75 54 70
To investigate the ternary system, a feed mixture close to the azeotropic point was chosen.
The composition of the azeotropic mixture was calculated with AP utilizing the Property model NRTL
for the vapour-liquid-liquid phases. At 1.01325 bar the azeotropic mixture contains 13.75% ethanol,
77.9% ethyl acetate and 8.35% water (mass fractions), boiling at 70.5 ◦ C. At the investigated feed
pressure of 5 bar, the boiling point of the azeotropic mixture is 122 ◦ C.
In order to test the scalability of the physico-chemical model large scale data from an industrial
pervaporation plant was compared to the respective simulation. The membrane module of the
pervaporation plant utilized flat-sheet membranes with a total area of 50 m2 . In total four of
these membrane modules were combined achieving a total membrane area of 200 m2 . The feed
stream is divided onto two modules (parallel array). After each of these modules the retentate
stream is heated-up to feed solution conditions and enters a second membrane module (series
array). This module configuration was adapted for the membrane module submodel. 15,000 kg
of ethanol/water feed solution were dehydrated for 24 h with the pervaporation plant with samples
taken every 2 h. The same membrane type was used for the industrial plant as in the experimental
pervaporation unit. The flow regime is comparable to the lab scale module. Table 5 list the set process
parameters utilized to generate the large-scale data.
Table 5. Process parameters and ethanol mass fraction after 24 h for the industrial pervaporation plant.
15,000 kg of ethanol/water solution were dehydrated, resulting in a total permeate mass of 1070 kg.
.
wstart
f,ethanol (%) Tf (◦ C) pf (bar) pp (mbar) Vf (L h−1 ) wend
f,ethanol (%)
4. Results
In all pervaporation experiments the mass balances were closed with lower than 7% deviation
(4.1% averaged). The reproducibility was examined by three centre point experiments for both binary
systems as well as the ternary system experiments at 95 ◦ C, which were also conducted three times.
The averaged relative standard deviation for ethanol/water (ethyl acetate/water) was 4.6% (6%) for
the water mass fraction in the feed and 13% (28%) for the total permeate flux respectively. For the
ternary system, the average relative standard deviation was 2.8% for the water mass fraction in
the feed and 9.2% for the total permeate flux. The averaged 95% confidence interval for analytics
(Karl-Fischer-titration, gas chromatography) was smaller than 1%.
Tables 2–5 list the dehydration experiments conducted in this study, their respective process
conditions and the mixture composition at the end of the experiments. Figures 6 and 7 depict the
mean feed mass fraction and total permeate flux over the experimental duration for the centre point
experiments of the binary systems ethanol/water and ethyl acetate/water, coupled with the respective
error bars representing the standard deviation. The solid and dashed lines represent the respective
simulations. Table 6 lists the permeance data determined from the experiments listed in Tables 2 and 3
for both binary systems. Equation (25) proved to suit in describing the water and ethanol permeance in
the binary system. The stronger dependency of the permeance on the water concentration in the feed
for the binary system ethyl acetate/water could not be described well with this approach. Thereby
mean feed mass fraction and total permeate flux over the experimental duration for the centre point
experiments of the binary systems ethanol/water and ethyl acetate/water, coupled with the respective
error bars representing the standard deviation. The solid and dashed lines represent the respective
simulations. Table 6 lists the permeance data determined from the experiments listed in Tables 2 and
3 for both binary systems. Equation (25) proved to suit in describing the water and ethanol permeance
Membranes 2018, 8, 4 16 of 27
in the binary system. The stronger dependency of the permeance on the water concentration in the
feed for the binary system ethyl acetate/water could not be described well with this approach.
Thereby (26)
Equation Equation (26) was
was tested and tested
reflectedand
thereflected
water and theethyl
water and permeance
acetate ethyl acetate permeance
well. Regarding well.
its
Regarding
physical its physical
properties, suchproperties,
as polaritysuch
andas polarity
size, waterand size, to
is closer water is closer
ethanol than to ethanol
to ethyl than to
acetate. ethyl
Hence,
acetate.
the Hence, the
dehydration dehydration
of the of the ethyl
ethyl acetate/water acetate/water
system system
has a higher has a higher
selectivity selectivity
than the than the
dehydration of
dehydration
ethanol. of ethanol.
In order In order
to adequately to adequately
describe describe
these differences in these differences
selectivity, in permeance
different selectivity,functions
different
permeance
can functions can be suitable.
be suitable.
Figure
Figure6.6. Average
Averagewater
watermass
massfraction
fractionof
ofthe
the feed
feed and
and total
total permeate
permeate flux
flux of
of the
the thrice-repeated
thrice-repeatedcentre
centre
point
point experiments over the experimental duration for the binary system ethanol/water. The errorbars
experiments over the experimental duration for the binary system ethanol/water. The error bars
represent
2018,the
represent
Membranes the
8, 4 standard
standarddeviation,
deviation,the
theline
linethe
thesimulation
simulationresult.
result. 16 of 27
Figure7.7. Average
Figure Averagewater
watermass
massfraction
fractionof
ofthe
thefeed
feedand
andtotal
total permeate
permeateflux
fluxof
ofthe
the thrice-repeated
thrice-repeatedcentre
centre
pointexperiments
point experimentsover
overthe
theexperimental
experimentalduration
durationforforthe
thebinary
binarysystem
systemethyl
ethylacetate/water.
acetate/water.The
Theerror
error
bars represent the standard deviation, the line the simulation result.
bars represent the standard deviation, the line the simulation result.
Table 6. Determined permeance data for the two binary systems by a nonlinear least squares
regression, minimizing the weighted square error between experiments and simulation. For the
simulation of the ternary system, the organic solvent permeance data was adopted from the binary
data. The water permeance data for the ternary simulation was adopted from the ethyl acetate/water data.
Table 6. Determined permeance data for the two binary systems by a nonlinear least squares regression,
minimizing the weighted square error between experiments and simulation. For the simulation of
the ternary system, the organic solvent permeance data was adopted from the binary data. The water
permeance data for the ternary simulation was adopted from the ethyl acetate/water data.
In order to validate the physico-chemical model, simulations based on the model coupled with
the determined permeance data were run and compared to a second set of binary experiments
listed in Tables 2 and 3 as well. Figures 8–10 depict the simulated and experimental water mass
fraction of the feed and permeate as well as the water flux over time for three exemplary ethanol/water
experiments. Figures 11–13 show the equivalent comparisons for three ethyl acetate/water experiments.
The simulations of the ternary system were conducted with the permeance data obtained from the
binary systems. For ethyl acetate and ethanol, the same correlation and parameters were used to
describe the binary and ternary system. To describe the water permeance of the ternary system the
water permeance data from the ethyl acetate/water experiments was utilized. Figures 14–16 show
the mass fraction of feed and permeate as well as the permeate flux of all three components over time
for the experiment at 95 ◦ C feed temperature. The average value of the thrice-conducted experiment
is thereby compared to the corresponding simulation. The ternary system was investigated for feed
temperatures of 95 ◦ C and 75 ◦ C. In Figure 17 the measured and simulated water mass fractions of
the feed progression for both temperatures are compared. This comparison is not only of interest
as an example of the temperature-dependency of pervaporation processes, but also to test if the
simulation reflects this dependency correctly. The scalability of the developed physico-chemical
model was investigated by comparing measured large-scale data to a simulation based on the model.
Figures 18–20 show the water mass fraction of the feed and permeate and the water flux progression
Membranes
over 24 h 2018, 8, 4
for process conditions listed in Table 5. 17 of 27
Figure 8.8.Experimental
Figure Experimentalwater
water mass
mass fractions
fractions (feed)(feed) overfor
over time time forexperiments
three three experiments used
used for the for the
validation
validation of the permeance function for ethanol/water compared with the corresponding
of the permeance function for ethanol/water compared with the corresponding simulation result. simulation
result.
Figure 8. Experimental water mass fractions (feed) over time for three experiments used for the
Figure 8. Experimental water mass fractions (feed) over time for three experiments used for the
validation
Membranes 2018, 8, 4of the permeance function for ethanol/water compared with the corresponding simulation 18 of 27
validation of the permeance function for ethanol/water compared with the corresponding simulation
result.
result.
Figure
Figure 9.Experimental
9. 9. Experimental water mass
watermass
mass fractions (permeate)
fractions over over
(permeate) time for three
time experiments
forexperiments used for the
three experiments used
Figure Experimental water fractions (permeate) over time for three used for the
validation
forvalidation of
the validation the permeance function for ethanol/water compared with the corresponding simulation
of the of the permeance
permeance functionfunction for ethanol/water
for ethanol/water compared
compared with with the corresponding
the corresponding simulation
result.
simulation
result. result.
Figure 10. Experimental water flux over time for three experiments used for the validation of the
Figure10.10.Experimental
Experimentalwater
waterflux
flux over
over time
time for
for three
three experiments used for
forthe
thevalidation of of
thethe
Figure
permeance function for ethanol/water compared with theexperiments used
corresponding simulation validation
result.
permeance function for ethanol/water compared with the corresponding simulation result.
permeance
Membranes 2018,function
8, 4 for ethanol/water compared with the corresponding simulation result. 18 of 27
Figure
Figure 11.11.Experimental
Experimentalwater
water mass
mass fractions
fractions (feed)
(feed) over
over time
timefor
forthree
threeexperiments
experimentsused forfor
used thethe
validation
validation of ofthethepermeance
permeancefunction
functionfor
for ethyl
ethyl acetate/water
acetate/water compared
comparedwith
withthe
thecorresponding
corresponding
simulation result.
simulation result.
Figure 11. Experimental water mass fractions (feed) over time for three experiments used for the
Figure 11. Experimental water mass fractions (feed) over time for three experiments used for the
validation of the permeance function for ethyl acetate/water compared with the corresponding
Membranes 2018, 8, 4 of the permeance function for ethyl acetate/water compared with the corresponding 19 of 27
validation
simulation result.
simulation result.
Figure 13. Experimental water flux over time for three experiments used for the validation of the
Figure13.13. Experimentalwater
Experimental waterflux
flux over
over time
time for
for three
three experiments
experiments used
usedfor
forthe
thevalidation
validationof the
Figure
permeance function for ethyl acetate/water compared with the corresponding simulation result. of the
permeance
permeance function for ethyl acetate/water compared with the corresponding simulation result.
Membranes 2018,function
8, 4 for ethyl acetate/water compared with the corresponding simulation result. 19 of 27
Figure 15. Experimental and simulation permeate mass fractions for the ternary system at 95 feed
Figure 15. Experimental
Experimental and simulation permeate mass fractions for the ternary system
◦ at 95 feed
Figure 15.
temperature °C. and simulation permeate mass fractions for the ternary system at 95 C feed temperature.
temperature °C.
Figure 16. Experimental and simulation permeate fluxes for the ternary system◦ at 95 °C feed
Figure Experimental
16.2018,
Figure
Membranes 16. and simulation
8, 4Experimental permeate
and simulation fluxes fluxes
permeate for thefor
ternary system at
the ternary 95 Catfeed
system 95 temperature.
°C feed
20 of 27
temperature.
temperature.
Figure 18. Comparison of the measured and simulated water mass fraction (feed) for the
Figure 18. Comparison
pervaporation plant.
Figure 18. Comparison of the measured
of the measured and water
and simulated simulated water mass
mass fraction (feed)fraction (feed) for the
for the pervaporation plant.
pervaporation plant.
FigureFigure
Membranes 19.
19. 2018, 8, 4Comparison
Comparison of the measured
of the measured andwater
and simulated simulated water mass
mass fraction fraction
(permeate) (permeate)
for the for the
pervaporation21plant.
of 27
Figure 19. Comparison
pervaporation plant. of the measured and simulated water mass fraction (permeate) for the
pervaporation plant.
5. Discussion
The focus of this work was the model development and validation. Hence, the variation in feed
composition, feed temperature, permeate pressure and feed velocity for both binary systems allowed
to test the derived model for a variation of process conditions. This variation also enables the
Membranes 2018, 8, 4 22 of 27
Figure 20. Comparison of the measured and simulated water flux for the pervaporation plant.
5. Discussion
The focus of this work was the model development and validation. Hence, the variation in feed
composition,
composition, feedfeed temperature,
temperature,permeate
permeatepressure
pressureandandfeed
feedvelocity
velocity for
for both
both binary
binary systems
systems allowed
allowed to
to test the derived model for a variation of process conditions. This variation also
test the derived model for a variation of process conditions. This variation also enables the investigation enables the
investigation
of the influence of ofthe influence
these processof these process
conditions on keyconditions
values such onaskey values such
the retentate as the retentate
composition, flux or
composition,Both
permeance. flux or permeance.
binary systems Both binarylarger
showed systems showedfluxes
permeate largerandpermeate fluxes and
permeances permeances
for larger water
for larger water concentrations in the feed. This effect is known in literature as
concentrations in the feed. This effect is known in literature as the coupling between water and the coupling between
water
organicand organic
solvent solvent permeances
permeances due to the
due to the swelling swelling
of the of themembrane
polymeric polymeric[2,26,27].
membrane [2,26,27].
Comparing
Comparing the binaryof
the binary selectivity selectivity of water/ethanol
water/ethanol to water/ethyl
to water/ethyl acetate
acetate in Figurein 21
Figure
two 21 two characteristics
characteristics strike
strike the eye. The selectivity of water/ethyl acetate is significantly larger than water/ethanol.
the eye. The selectivity of water/ethyl acetate is significantly larger than water/ethanol. Due to Due to the
higher polarity of
higher polarity of ethanol
ethanol compared
compared to to ethyl
ethyl acetate
acetate this
this observation
observation was
was expectable.
expectable. Both
Both selectivities
selectivities
drop
drop with higher water content, due to the mentioned coupling. The drop in selectivity seems to
with higher water content, due to the mentioned coupling. The drop in selectivity seems be
to be
proportionally with
proportionally with similar
similar observations
observationsmade madein inliterature
literaturefor
forother
othersolvents
solvents[2].
[2].
Figure 21.
21. Binary water/ethanol
water/ethanol selectivity
selectivity (blue
(blue squares) and water/ethyl
water/ethyl acetate
acetate selectivity
selectivity (red
circles plotted over the water mass fraction (feed) for the ternary system. The error bars represent the
standard deviation.
For the
For the varied
varied temperature range aa difference
temperature range difference in
in the
the permeate
permeate flux
flux but
but not
not in
in the
the permeance
permeance was
was
detected. The permeance parameter is thereby 0 for all components. Hence, the
detected. The permeance parameter bi is thereby 0 for all components. Hence, the temperature-dependency
of the permeate flux is only considered in the physico-chemical model by the driving force. An example
for the validity of this approach is seen in Figure 17. The determined permeance data is able to equally
describe the water mass fraction in the feed tank for 75 ◦ C as for 95 ◦ C. Another example is, for instance,
the shown trajectories of experiments 3, 6 and 8 in Figures 8–10 for ethanol/water.
The combination of feed water content and temperature can in each case be very well described.
The by theory anticipated influence of the permeate pressure on the permeate flux and thereby the
dehydration of the feed mixture was well observed, for example by comparing exp. 10 and 11 in
Table 2. Out of the four varied process conditions, the feed volume flow and thereby the feed velocity
showed the smallest impact on the permeate flux and mass fractions for the chosen range.
Figures 8–13 illustrate exemplary via the respective comparison between experimental and
simulated data the validity of the developed physico-chemical model coupled with the permeance data
listed in Table 6. The temperature-independency of the permeance discussed above was again found by
the nonlinear regression done for the determination of the permeance data, resulting in the parameter
bq,i being 0 for water and both organic solvents. As addressed in Section 2.1.2 a temperature-insensitive
permeance or even lower permeances for higher temperatures are possible.
Membranes 2018, 8, 4 23 of 27
The investigation of the ternary mixture for two temperatures with a feed composition near the
industrially-relevant azeotropic mixture revealed two findings. First of all, the temperature-independency
of the water and organics permeances was found as well (Figure 17). Second of all utilizing the
permeance data of the organic solvents obtained from the binary experiments in the simulation of the
ternary system resulted in a very good agreement between simulated and measured data. In case
of the water permeances lead only the water permeance parameters obtained by the binary ethyl
acetate/water experiments to satisfactory simulation results. Utilization of the discussed permeance
data resulted overall in the very good agreement between the experiments and the simulation,
illustrated throughout Figures 14–16. The deviation of the experimental data at 6 h in Figure 15
is in the author’s opinion an experimental error. The applicability of the binary permeance data
for the simulation of the ternary system is a very interesting result. The feed mixture of the ternary
experiments contains about 4.75 times the amount of ethyl acetate than ethanol. In the author’s opinion,
this is a very probable reason for the better applicability of the water permeance parameters from
the ethyl acetate/water data. For a more detailed insight into the permeances of ternary mixtures
compared to binary mixtures, an investigation through a detailed experimental design for ternary
mixtures incorporating a more profound theory of the sorption and diffusion mechanisms [28,29]
would be an appealing topic for future work, but was beyond the scope of this study.
In order to apply the developed physico-chemical model as an engineering tool for the design
of pervaporation processes, a proof of principle for the scalability of the model is of key importance.
For this reason, large-scale data of a pervaporation unit with 200 m2 membrane area was compared
to a simulation run with the model. Since the large-scale data was generated for the dehydration
of ethanol/water with an identical membrane than in the experiments conducted for this study,
the permeance data listed in Table 6 could be transferred. The measured and simulated water
concentration in feed and permeate, as well as the permeate flux is shown in Figures 18–20.
The simulation of the industrial-scale pervaporation unit displays a very good accordance to the
measured data. The temperature drops on the feed side of the membrane due to vaporization of the
permeating components has a significantly larger effect than temperature polarization. As the key
accomplishment of this work the scalability of the developed physico-chemical model combined with
laboratory/mini-plant experiments are thereby proven.
Further, the parity plots between the simulation and measured data displayed in Figure 22 for
the water mass fraction in the feed, in Figure 23 for the water mass fraction in the permeate and in
Figure 24 for the water flux demonstrate the very good accordance not only for the industrial scale but
also for the
Membranes tested
2018, 8, 4 ternary mixture. 23 of 27
Figure22.
Figure 22. Parity
Parity plot
plot between
between the
the experimental
experimental and
and simulated
simulated water
water mass
mass fraction
fraction (feed)
(feed) for
forthe
the
ternary system at 95◦°C feed temperature.
ternary system at 95 C feed temperature.
Figure
Figure 22.
22.8,Parity
Parity plot
plot between
between the
the experimental
experimental and
and simulated
simulated water
water mass
mass fraction
fraction (feed)
(feed) for
for the
the
Membranes 2018, 4 24 of 27
ternary
ternary system
system at
at 95
95 °C
°C feed
feed temperature.
temperature.
Figure 23.
Figure 23. Parity
Parity plot
plot between
between the
the experimental
experimental and
and simulated
simulated water
water mass
mass fraction
fraction (permeate)
(permeate) for
(permeate) for the
for the
ternary system
ternary system at
at 95
95 ◦°C
C feed
feed temperature.
temperature.
system at 95 °C
Figure 24.
Figure 24. Parity
Parity plot between
between the experimental
experimental and
and simulated
simulated water
water flux
flux for
for the
the ternary
ternary system
system at
at 95
95 °C
°C
Figure 24. Parityplot
plot betweenthe
the experimental and simulated water flux for the ternary system at
feed
feed temperature.
◦ Ctemperature.
95 feed temperature.
6. Conclusions
This study shows the development of a physico-chemical model for pervaporation processes.
The solution-diffusion mechanism, concentration and temperature polarization, axial dispersion,
pressure drop and the temperature gradient due to vaporization of the permeate were implemented
into the model. Dehydration experiments with the binary systems ethanol/water and ethyl acetate
water were conducted in order to determine the three permeance function parameters needed for
each component. Other relevant model parameters were either directly calculated or determined by
correlations. The developed model combined with the determined permeance data was validated
with a second set of experiments. Dehydration experiments were conducted with the industrially
relevant ternary mixture ethanol/ethyl acetate/water. The simulation of the ternary mixture with the
permeance data received from the binary experiments (the water permeance was implemented from the
ethyl acetate/water data) showed a very good agreement with the experimental data. The scalability
of the developed model was investigated by comparing measured data from an industrial-scale
pervaporation unit (200 m2 membrane area, 15,000 kg ethanol/water feed) with the simulation
utilizing the permeance based on the mini-plant experiments. The very good agreement between the
measured and simulated permeate flux, feed and permeate composition data proved the scalability of
correlations. The developed model combined with the determined permeance data was validated
waterawere
with second conducted in order to Dehydration
set of experiments. determine theexperiments
three permeance function parameters
were conducted needed for
with the industrially
with a second set of experiments. Dehydration experiments were conducted with the industrially
each component. Other relevant model parameters were either directly calculated
relevant ternary mixture ethanol/ethyl acetate/water. The simulation of the ternary mixture with the or determined by
relevant ternary mixture ethanol/ethyl acetate/water. The simulation of the ternary mixture with the
correlations.data
permeance Thereceived
developedfrommodel combined
the binary with the(the
experiments determined permeance
water permeance wasdata was validated
implemented from
permeance data received from the binary experiments (the water permeance was implemented from
withethyl
the a second set of experiments.
acetate/water data) showed Dehydration experiments
a very good agreementwerewith
conducted with the industrially
the experimental data. The
the ethyl acetate/water data) showed a very good agreement with the experimental data. The
relevant ternary
scalability
Membranes mixture
of the developed
8, 4 ethanol/ethyl acetate/water.
model was investigated The simulation
by comparing of the
measured ternary
data from anmixture with25the
industrial-scale
of 27
scalability of2018,
the developed model was investigated by comparing measured data from an industrial-scale
permeance dataunit
pervaporation received
(200 from the binary area,
m22 membrane experiments
15,000 (the water permeance
kg ethanol/water was
feed) implemented
with from
the simulation
pervaporation unit (200 m membrane area, 15,000 kg ethanol/water feed) with the simulation
the ethylthe
utilizing acetate/water
permeance based data) onshowed a very good
the mini-plant agreement
experiments. The with the experimental
very good data. The
agreement between the
utilizing
the the permeance
model. Thereby, thebased on the mini-plant
physico-chemical model experiments.
combined Theexperiments
with very good agreement
on a between
mini-plant thecan
scale
scalability of
measured andthesimulated
developedpermeate
model was investigated
flux, feed andby comparing
permeate measured data
composition datafrom an industrial-scale
proved the scalability
measured
be and simulated permeate forflux, feed and permeate composition data proved the scalability
of theused
pervaporation
model.as an engineering
unit
Thereby, (200
them 2 tool
membrane the design
physico-chemical area, of pervaporation
15,000
model processes.
kg ethanol/water
combined with feed) with
experiments the simulation
on a mini-plant scale
of the model. Thereby, the physico-chemical model combined with experiments on a mini-plant scale
utilizing
can the as
be used permeance based tool
an engineering on theformini-plant
the designexperiments.
of pervaporationThe very good agreement between the
processes.
canAcknowledgments: The authors
be used as an engineering thank
tool the ITVP
for the designlabofteam, especially Frank
pervaporation Steinhäuser, Martina Ketterer and
processes.
measured and simulated
Volker Strohmeyer permeate
for their efforts andflux, feed and permeate composition data proved the scalability
support.
Abbreviations
of Author
the model. Thereby, the physico-chemical model combined with experiments on a mini-plant scale
Contributions: Holger Thiess conceived and designed the experiment as well as wrote the paper.
Abbreviations
canHolger
be used
Symbols as an
Thiess engineering
performed tool for experiments.
the described the design ofAxelpervaporation
Schmidt and processes.
Jochen Strube substantively revised the
Symbols
work
A
and contributed the materials. Jochen Strube is responsible for conception and supervision.
m 2 area
A
Conflicts ofm
Abbreviations 2 area
− Interest: The authors declare no
permeance conflict of interest.
coefficient
− permeance coefficient
Symbols −− activity
activity
Abbreviations
−2 Sh/Nu correlation parameters
A −m area correlation
Sh/Nu parameters
− permeance coefficient
Symbols −− permeance coefficient
permeance coefficient
A − m2 activity area
concentration
³ concentration
Ai − −³ Sh/Nu correlation
permeance parameters
coefficient
− specific
permeanceheat capacity
coefficient
a − activity
specific heat capacity
a1 to a4 − Sh/Nu correlation parameters
concentration
diffusion coefficient
Bi −³ diffusion permeance
coefficient coefficient
c mol or kg diameter
specific heat capacity
concentration
m3 m3 diameter
~ kJ kJ enthalpy flow
cP kg K or kmolenthalpy
K
diffusion
specific heat capacity
coefficient
flow
D m2 diffusion coefficient
s specific
d m diameterenthalpy
diameter
specific enthalpy
kJ height
s
enthalpy flow
enthalpy flow
~ kJ
height
h kg permeate specific
flux enthalpy
specific enthalpy
permeate flux
h − m height factor
proportionality
J − kg
or mol proportionality
height permeate factor
flux
m h2 2
m h length
length
L − permeate proportionality
flux
sorption coefficient
factor
l ³⋅ m sorption coefficient
length
−³ ⋅ proportionality factor
K / kg mass transfer coefficient
sorption coefficient
/ 3
m ·bar mass
lengthtransfer coefficient
k m/h molar mass mass transfer coefficient
kg molar mass
sorption coefficient
M ³⋅ kmol mass molar mass
m / kg mass transfer
masscoefficient
kg mass flow
s
mass
molarflow
massmass flow
kmol
s mole flow
mass flowmole flow
Membranes 2018, 8, 4 mole 25 of 27
Nu − − Nusselt Nusselt number
number
− kg m mass flow
Nusselt number
P m2 h bar
permeability
permeability
p bar permeance
mole flowpressure
permeability
pressure
Pr − Prandtl number
−
−
pressure
Nusselt number
permeancenumber
coefficient
kg Prandtl
Q − m2 h bar Prandtl permeance
number
Q0i kg permeability
permeance coefficient
m2 h bar gas constant
kJ pressure gas constant
R
− mol K Reynolds number
Re − − Prandtl number
Reynolds number
− Schmidt number
Sc − − SherwoodSchmidt
numbernumber
Sh − temperatureSherwood number
T K temperature
m velocity
u s velocity
Uwetted m wetted perimeter
wetted perimeter
L volume flow
h volume flow
v m3 molar volume
³ mol
kg molar volume
wi kg mass fraction
w m width
mass fraction
kmol
x kmol molar fraction (liquid)
kmol width
y kmol molar fraction (gaseous)
molar fraction (liquid)
Greek letters
J
α hK heat transfer coefficient
γ − activity coefficient
δ m boundary layer thickness
ζ − drag coefficient
W
λ mK thermal conductivity
η Pa·s dynamic viscosity
kJ
µ mol
chemical potential
ν m2 kinematic viscosity
s
kg
ρ m3
density
Subscripts
b bulk
ch channel
Diff diffusion
eff effective
f feed
fm membrane—feed side
G gas
h hydraulic
i,j components
L liquid
l length
m membrane
p permeate
pm membrane—permeate side
Q cross-section
r retentate
sat saturated
tot total
TM transmembrane
vap vaporization
0 reference
References
1. Stichlmair, J.G.; Fair, J.R. Distillation: Principles and Practices; WILEY-VCH Verlag Press: New York, NY,
USA, 1998.
2. Koch, K.; Górak, A. Pervaporation of binary and ternary mixtures of acetone, isopropyl alcohol and water
using polymeric membranes: Experimental characterisation and modelling. Chem. Eng. Sci. 2014, 115, 95–114.
[CrossRef]
3. Hessel, V.; Gürsel, I.V.; Wang, Q.; Noël, T.; Lang, J. Potenzialanalyse von milli-und mikroprozesstechniken
für die verkürzung von prozessentwicklungszeiten—Chemie und prozessdesign als intensivierungsfelder:
Potential analysis of smart flow processing and micro process technology for fastening process
development—Use of chemistry and process design as intensification fields. Chem. Ing. Tech. 2012, 84,
660–684.
4. Del Pozo Gómez, M.T.; Carreira, P.R.; Repke, J.-U.; Klein, A.; Brinkmann, T.; Wozny, G. Study of a novel heat
integrated hybrid pervaporation distillation process: Simulation and experiments. Comput. Aided Process Eng.
2008, 25, 73–78.
5. Wijmans, J.G.; Baker, R.W. The solution-diffusion model: A review. J. Membr. Sci. 1995, 107, 1–21. [CrossRef]
6. Wijmans, J.G.; Baker, R.W. A simple predictive treatment of the permeation process in pervaporation.
J. Membr. Sci. 1993, 79, 101–113. [CrossRef]
7. Kreis, P.; Górak, A. Process Analysis of hybrid separation processes. Chem. Eng. Res. Des. 2006, 84, 595–600.
[CrossRef]
Membranes 2018, 8, 4 27 of 27
8. Schiffmann, P.; Repke, J.-U. Design of pervaporation modules based on computational process modelling.
In Proceedings of the 21st European Symposium on Computer Aided Process Engineering, Chalkidiki,
Greece, 29 May–2 June 2011.
9. Koch, K.; Sudhoff, D.; Kreiß, S.; Górak, A.; Kreis, P. Optimisation-based design method for membrane-assisted
separation processes. Chem. Eng. Process. Process Intensif. 2013, 67, 2–15. [CrossRef]
10. Wijmans, J.G. Process performance = membrane properties + operating conditions. J. Membr. Sci. 2003, 220,
1–3. [CrossRef]
11. Baker, R.W.; Wijmans, J.G.; Athayde, A.L.; Daniels, R.; Ly, J.H.; Le, M. The effect of concentration polarization
on the separation of volatile organic compounds from water by pervaporation. J. Membr. Sci. 1997, 137,
159–172. [CrossRef]
12. Leveque, M.A. Les Lois de la Transmission de Chaleur par Convection. Ann. Mines 1928, 13, 201–299.
13. Sieder, E.N.; Tate, G.E. Heat transfer and pressure drop of liquids in tubes. Ind. Eng. Chem. 1936, 28,
1429–1435. [CrossRef]
14. Gröber, H.; Erk, S. Fundamentals of Heat and Mass Transfer; McGraw-Hill Press: New York, NY, USA, 1961.
15. Hartnett, J.P.; Mickley, H.S.; Eckert, E.R.G. Recent Advances in Heat and Mass Transfer; McGraw-Hill Press:
New York, NY, USA, 1961.
16. Thiess, H.; Leuthold, M.; Grummert, U.; Strube, J. Module design for ultrafiltration in biotechnology:
Hydraulic analysis and statistical modelling. J. Membr. Sci. 2017, 540, 440–453. [CrossRef]
17. Wijmans, J.G.; Athayde, A.L.; Daniels, R.; Ly, J.H.; Kamaruddin, H.D.; Pinnau, I. The role of boundary layers
in the removal of volatile organic compounds from water by pervaporation. J. Membr. Sci. 1996, 109, 135–146.
[CrossRef]
18. Melin, T.; Rautenbach, R. Membranverfahren. Grundlagen der Modul- und Anlagenauslegung; Springer-Verlag
Press: Berlin, Germany, 2007.
19. Levenspiel, O. Chemical Reaction Engineering, 3rd ed.; John Wiley & Sons Press: New York, NY, USA, 1999.
20. Lipnizki, F.; Olsson, J.; Trägårdh, G. Scale-up of pervaporation for the recovery of natural aroma compounds
in the food industry. Part 1: Simulation and performance. J. Food Eng. 2002, 54, 183–195. [CrossRef]
21. Lipnizki, F.; Field, R.W. Simulation and process design of pervaporation plate-and-frame modules to recover
organic compounds from waste water. Chem. Eng. Res. Des. 1999, 77, 231–240. [CrossRef]
22. Vallieres, C.; Favre, E. Vacuum versus sweeping gas operation for binary mixtures separation by dense
membrane processes. J. Membr. Sci. 2004, 244, 17–23. [CrossRef]
23. Grote, F.; Fröhlich, H.; Strube, J. Integration of reverse-osmosis unit operations in biotechnology process
design. Chem. Eng. Technol. 2012, 35, 191–197. [CrossRef]
24. Renon, H.; Prausnitz, J.M. Estimation of parameters for the NRTL equation for excess gibbs energies of
strongly nonideal liquid mixtures: Industrial & engineering chemistry process design and development.
Ind. Eng. Chem. Proc. Des. Dev. 1969, 8, 413–419.
25. Dennis, J.E.; Gay, D.M.; Welsch, R.E. An adaptive nonlinear least-squares algorithm. ACM Trans. Math. Softw.
1981, 7, 348–368. [CrossRef]
26. Baker, R.W. Membrane Technology and Applications; John Wiley & Sons Ltd.: Chichester, UK, 2012.
27. Heintz, A.; Stephan, W. A generalized solution—Diffusion model of the pervaporation process through
composite membranes Part II. Concentration polarization, coupled diffusion and the influence of the porous
support layer. J. Membr. Sci. 1994, 89, 153–169. [CrossRef]
28. Heintz, A.; Stephan, W. A generalized solution—Diffusion model of the pervaporation process through
composite membranes Part I. Prediction of mixture solubilities in the dense active layer using the UNIQUAC
model. J. Membr. Sci. 1994, 89, 143–151. [CrossRef]
29. Krishna, R. The Maxwell-Stefan description of mixture diffusion in nanoporous crystalline materials.
Microporous Mesoporous Mater. 2014, 185, 30–50. [CrossRef]
© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).